Related
This question already has answers here:
'for' loop vs vectorization in MATLAB
(5 answers)
Closed 3 years ago.
In Matlab, I am trying to vectorise my code to improve the simulation time. However, the result I got was that I deteriorated the overall efficiency.
To understand the phenomenon I created 3 distinct functions that does the same thing but with different approach :
The main file :
clc,
clear,
n = 10000;
Value = cumsum(ones(1,n));
NbLoop = 10000;
time01 = zeros(1,NbLoop);
time02 = zeros(1,NbLoop);
time03 = zeros(1,NbLoop);
for test = 1 : NbLoop
tic
vector1 = function01(n,Value);
time01(test) = toc ;
tic
vector2 = function02(n,Value);
time02(test) = toc ;
tic
vector3 = function03(n,Value);
time03(test) = toc ;
end
figure(1)
hold on
plot( time01, 'b')
plot( time02, 'g')
plot( time03, 'r')
The function 01:
function vector = function01(n,Value)
vector = zeros( 2*n,1);
for k = 1:n
vector(2*k -1) = Value(k);
vector(2*k) = Value(k);
end
end
The function 02:
function vector = function02(n,Value)
vector = zeros( 2*n,1);
vector(1:2:2*n) = Value;
vector(2:2:2*n) = Value;
end
The function 03:
function vector = function03(n,Value)
MatrixTmp = transpose([Value(:), Value(:)]);
vector = MatrixTmp (:);
end
The blue plot correspond to the for - loop.
n = 100:
n = 10000:
When I run the code with n = 100, the more efficient solution is the first function with the for loop.
When n = 10000 The first function become the less efficient.
Do you have a way to know how and when to properly replace a for-loop by a vectorised counterpart?
What is the impact of index searching with array of tremendous dimensions ?
Does Matlab compute in a different manner an array of dimensions 3 or higher than a array of dimension 1 or 2?
Is there a clever way to replace a while loop that use the result of an iteration for the next iteration?
Using MATLAB Online I see something different:
n 10000 100
function01 5.6248e-05 2.2246e-06
function02 1.7748e-05 1.9491e-06
function03 2.7748e-05 1.2278e-06
function04 1.1056e-05 7.3390e-07 (my version, see below)
Thus, the loop version is always slowest. Method #2 is faster for very large matrices, Method #3 is faster for very small matrices.
The reason is that method #3 makes 2 copies of the data (transpose or a matrix incurs a copy), and that is bad if there's a lot of data. Method #2 uses indexing, which is expensive, but not as expensive as copying lots of data twice.
I would suggest this function instead (Method #4), which transposes only vectors (which is essentially free). It is a simple modification of your Method #3:
function vector = function04(n,Value)
vector = [Value(:).'; Value(:).'];
vector = vector(:);
end
Do you have a way to know how and when to properly replace a for-loop by a vectorised counterpart?
In general, vectorized code is always faster if there are no large intermediate matrices. For small data you can vectorize more aggressively, for large data sometimes loops are more efficient because of the reduced memory pressure. It depends on what is needed for vectorization.
What is the impact of index searching with array of tremendous dimensions?
This refers to operations such as d = data(data==0). Much like everything else, this is efficient for small data and less so for large data, because data==0 is an intermediate array of the same size as data.
Does Matlab compute in a different manner an array of dimensions 3 or higher than a array of dimension 1 or 2?
No, not in general. Functions such as sum are implemented in a dimensionality-independent waycitation needed.
Is there a clever way to replace a while loop that use the result of an iteration for the next iteration?
It depends very much on what the operations are. Functions such as cumsum can often be used to vectorize this type of code, but not always.
This is my timing code, I hope it shows how to properly use timeit:
%n = 10000;
n = 100;
Value = cumsum(ones(1,n));
vector1 = function01(n,Value);
vector2 = function02(n,Value);
vector3 = function03(n,Value);
vector4 = function04(n,Value);
assert(isequal(vector1,vector2))
assert(isequal(vector1,vector3))
assert(isequal(vector1,vector4))
timeit(#()function01(n,Value))
timeit(#()function02(n,Value))
timeit(#()function03(n,Value))
timeit(#()function04(n,Value))
function vector = function01(n,Value)
vector = zeros(2*n,1);
for k = 1:n
vector(2*k-1) = Value(k);
vector(2*k) = Value(k);
end
end
function vector = function02(n,Value)
vector = zeros(2*n,1);
vector(1:2:2*n) = Value;
vector(2:2:2*n) = Value;
end
function vector = function03(n,Value)
MatrixTmp = transpose([Value(:), Value(:)]);
vector = MatrixTmp(:);
end
function vector = function04(n,Value)
vector = [Value(:).'; Value(:).'];
vector = vector(:);
end
I have tried searching for an answer on this, but can't find one that specifically addresses my issue. Although vectorizing in MATLAB must draw many questions in, the problem I am having is less general than typical examples I have found on the web. My background is more in C++ than MATLAB, so this is an odd concept for me to get my head around.
I am trying to evolve a Hamiltonian matrix from it's initial state (being a column vector where all elements but the last is a 0, and the last is a 1) to a final state as time increases. This is achieved by sequentially applying a time evolution operator U to the state. I also want to use the new state at each time interval to calculate an observable property.
I have achieved this, as can be seen in the code below. However, I need to make this code as efficient as possible, and so I was hoping to vectorize, rather than rely on for loops. However, I am unsure of how to vectorize this code. The problem I have is that on each iteration of the for loop, the column vector psi should change its values. Each new psi is then used to calculate my observable M for each interval of time. I am unsure of how to track the evolution of psi such that I can end up with a row vector for M, giving the outcome of the application of each new psi.
time = tmin:dt:tmax;
H = magic(2^N)
X = [0,1;1,0]
%%% INITIALISE COLUMN VECTOR
init = sparse(2^N,1);
init(2^N) = 1;
%%% UNITARY TIME EVOLUTION OPERATOR
U = expm(-1i*H*dt);
%%% TIME EVOLVUTION
for num = 1:length(time)
psi = U*init;
init = psi;
%%% CALCULATE OBSERVABLE
M(num) = psi' * kron(X,speye(2^(N-1))) * psi
end
Any help would be greatly appreciated.
I have quickly come up with the following partially vectorized code:
time = tmin:dt:tmax;
H = magic(2^N);
X = [0,1;1,0];
%%% INITIALISE COLUMN VECTOR
init = sparse(2^N,1);
init(2^N) = 1;
%%% UNITARY TIME EVOLUTION OPERATOR
U = expm(-1i*H*dt);
%%% TIME EVOLVUTION
% preallocate psi
psi = complex(zeros(2^N, length(time)));
% compute psi for all timesteps
psi(:,1) = U*init;
for num = 2:length(time)
psi(:,num) = U*psi(:, num-1);
end
% precompute kronecker product (if X is constant through time)
F = kron(X,speye(2^(N-1)));
%%% CALCULATE OBSERVABLE
M = sum((psi' * F) .* psi.', 2);
However, it seems that the most computationally intensive part of your problem is computation of the psi. For that I can't see any obvious way to vectorize as it depends on the value computed in the previous step.
This line:
M = sum((psi' * F) .* psi.', 2);
is a little Matlab trick to compute psi(:,i)'*F*psi(:,i) in a vectorized way.
Summary: This question deals with the improvement of an algorithm for the computation of linear regression.
I have a 3D (dlMAT) array representing monochrome photographs of the same scene taken at different exposure times (the vector IT) . Mathematically, every vector along the 3rd dimension of dlMAT represents a separate linear regression problem that needs to be solved. The equation whose coefficients need to be estimated is of the form:
DL = R*IT^P, where DL and IT are obtained experimentally and R and P must be estimated.
The above equation can be transformed into a simple linear model after applying a logarithm:
log(DL) = log(R) + P*log(IT) => y = a + b*x
Presented below is the most "naive" way to solve this system of equations, which essentially involves iterating over all "3rd dimension vectors" and fitting a polynomial of order 1 to (IT,DL(ind1,ind2,:):
%// Define some nominal values:
R = 0.3;
IT = 600:600:3000;
P = 0.97;
%// Impose some believable spatial variations:
pMAT = 0.01*randn(3)+P;
rMAT = 0.1*randn(3)+R;
%// Generate "fake" observation data:
dlMAT = bsxfun(#times,rMAT,bsxfun(#power,permute(IT,[3,1,2]),pMAT));
%// Regression:
sol = cell(size(rMAT)); %// preallocation
for ind1 = 1:size(dlMAT,1)
for ind2 = 1:size(dlMAT,2)
sol{ind1,ind2} = polyfit(log(IT(:)),log(squeeze(dlMAT(ind1,ind2,:))),1);
end
end
fittedP = cellfun(#(x)x(1),sol); %// Estimate of pMAT
fittedR = cellfun(#(x)exp(x(2)),sol); %// Estimate of rMAT
The above approach seems like a good candidate for vectorization, since it does not utilize MATLAB's main strength that is MATrix operations. For this reason, it does not scale very well and takes much longer to execute than I think it should.
There exist alternative ways to perform this computation based on matrix division, as demonstrated here and here, which involve something like this:
sol = [ones(size(x)),log(x)]\log(y);
That is, appending a vector of 1s to the observations, followed by mldivide to solve the equation system.
The main challenge I'm facing is how to adapt my data to the algorithm (or vice versa).
Question #1: How can the matrix-division-based solution be extended to solve the problem presented above (and potentially replace the loops I am using)?
Question #2 (bonus): What is the principle behind this matrix-division-based solution?
The secret ingredient behind the solution that includes matrix division is the Vandermonde matrix. The question discusses a linear problem (linear regression), and those can always be formulated as a matrix problem, which \ (mldivide) can solve in a mean-square error sense‡. Such an algorithm, solving a similar problem, is demonstrated and explained in this answer.
Below is benchmarking code that compares the original solution with two alternatives suggested in chat1, 2 :
function regressionBenchmark(numEl)
clc
if nargin<1, numEl=10; end
%// Define some nominal values:
R = 5;
IT = 600:600:3000;
P = 0.97;
%// Impose some believable spatial variations:
pMAT = 0.01*randn(numEl)+P;
rMAT = 0.1*randn(numEl)+R;
%// Generate "fake" measurement data using the relation "DL = R*IT.^P"
dlMAT = bsxfun(#times,rMAT,bsxfun(#power,permute(IT,[3,1,2]),pMAT));
%% // Method1: loops + polyval
disp('-------------------------------Method 1: loops + polyval')
tic; [fR,fP] = method1(IT,dlMAT); toc;
fprintf(1,'Regression performance:\nR: %d\nP: %d\n',norm(fR-rMAT,1),norm(fP-pMAT,1));
%% // Method2: loops + Vandermonde
disp('-------------------------------Method 2: loops + Vandermonde')
tic; [fR,fP] = method2(IT,dlMAT); toc;
fprintf(1,'Regression performance:\nR: %d\nP: %d\n',norm(fR-rMAT,1),norm(fP-pMAT,1));
%% // Method3: vectorized Vandermonde
disp('-------------------------------Method 3: vectorized Vandermonde')
tic; [fR,fP] = method3(IT,dlMAT); toc;
fprintf(1,'Regression performance:\nR: %d\nP: %d\n',norm(fR-rMAT,1),norm(fP-pMAT,1));
function [fittedR,fittedP] = method1(IT,dlMAT)
sol = cell(size(dlMAT,1),size(dlMAT,2));
for ind1 = 1:size(dlMAT,1)
for ind2 = 1:size(dlMAT,2)
sol{ind1,ind2} = polyfit(log(IT(:)),log(squeeze(dlMAT(ind1,ind2,:))),1);
end
end
fittedR = cellfun(#(x)exp(x(2)),sol);
fittedP = cellfun(#(x)x(1),sol);
function [fittedR,fittedP] = method2(IT,dlMAT)
sol = cell(size(dlMAT,1),size(dlMAT,2));
for ind1 = 1:size(dlMAT,1)
for ind2 = 1:size(dlMAT,2)
sol{ind1,ind2} = flipud([ones(numel(IT),1) log(IT(:))]\log(squeeze(dlMAT(ind1,ind2,:)))).'; %'
end
end
fittedR = cellfun(#(x)exp(x(2)),sol);
fittedP = cellfun(#(x)x(1),sol);
function [fittedR,fittedP] = method3(IT,dlMAT)
N = 1; %// Degree of polynomial
VM = bsxfun(#power, log(IT(:)), 0:N); %// Vandermonde matrix
result = fliplr((VM\log(reshape(dlMAT,[],size(dlMAT,3)).')).');
%// Compressed version:
%// result = fliplr(([ones(numel(IT),1) log(IT(:))]\log(reshape(dlMAT,[],size(dlMAT,3)).')).');
fittedR = exp(real(reshape(result(:,2),size(dlMAT,1),size(dlMAT,2))));
fittedP = real(reshape(result(:,1),size(dlMAT,1),size(dlMAT,2)));
The reason why method 2 can be vectorized into method 3 is essentially that matrix multiplication can be separated by the columns of the second matrix. If A*B produces matrix X, then by definition A*B(:,n) gives X(:,n) for any n. Moving A to the right-hand side with mldivide, this means that the divisions A\X(:,n) can be done in one go for all n with A\X. The same holds for an overdetermined system (linear regression problem), in which there is no exact solution in general, and mldivide finds the matrix that minimizes the mean-square error. In this case too, the operations A\X(:,n) (method 2) can be done in one go for all n with A\X (method 3).
The implications of improving the algorithm when increasing the size of dlMAT can be seen below:
For the case of 500*500 (or 2.5E5) elements, the speedup from Method 1 to Method 3 is about x3500!
It is also interesting to observe the output of profile (here, for the case of 500*500):
Method 1
Method 2
Method 3
From the above it is seen that rearranging the elements via squeeze and flipud takes up about half (!) of the runtime of Method 2. It is also seen that some time is lost on the conversion of the solution from cells to matrices.
Since the 3rd solution avoids all of these pitfalls, as well as the loops altogether (which mostly means re-evaluation of the script on every iteration) - it unsurprisingly results in a considerable speedup.
Notes:
There was very little difference between the "compressed" and the "explicit" versions of Method 3 in favor of the "explicit" version. For this reason it was not included in the comparison.
A solution was attempted where the inputs to Method 3 were gpuArray-ed. This did not provide improved performance (and even somewhat degradaed them), possibly due to wrong implementation, or the overhead associated with copying matrices back and forth between RAM and VRAM.
I need the "for" loop in the following representative section of code to run as efficiently as possible. The mean function in the code is acting as a representative placeholder for my own function.
x = linspace(-1,1,15);
y = linspace(2,4,15);
[xgrid, ygrid] = meshgrid(x,y);
mc = rand(100000,1);
z=zeros(size(xgrid));
for i=1:length(xgrid)
for j=1:length(ygrid)
z(i,j) = mean(xgrid(i,j) + ygrid(i,j) + xgrid(i,j)*ygrid(i,j)*mc);
end
end
I have vectorized the code and improved its speed by about 2.5 times by building a matrix in which mc is replicated for each grid point. My implementation results in a very large matrix (3 x 22500000) filled with repeated data. I've mitigated the memory penalty of this approach by converting the matrix to single precision, but it seems like there should be a more efficient way to do what I want that avoids replicating so much data.
You could use bsxfun with few reshapes -
A = bsxfun(#times,y,x.'); %//'
B = bsxfun(#plus,y,x.'); %//'
C = mean(bsxfun(#plus,bsxfun(#times,mc,reshape(A,1,[])) , reshape(B,1,[])),1);
z_out = reshape(C,numel(x),[]).';
I have a MATLAB routine with one rather obvious bottleneck. I've profiled the function, with the result that 2/3 of the computing time is used in the function levels:
The function levels takes a matrix of floats and splits each column into nLevels buckets, returning a matrix of the same size as the input, with each entry replaced by the number of the bucket it falls into.
To do this I use the quantile function to get the bucket limits, and a loop to assign the entries to buckets. Here's my implementation:
function [Y q] = levels(X,nLevels)
% "Assign each of the elements of X to an integer-valued level"
p = linspace(0, 1.0, nLevels+1);
q = quantile(X,p);
if isvector(q)
q=transpose(q);
end
Y = zeros(size(X));
for i = 1:nLevels
% "The variables g and l indicate the entries that are respectively greater than
% or less than the relevant bucket limits. The line Y(g & l) = i is assigning the
% value i to any element that falls in this bucket."
if i ~= nLevels % "The default; doesnt include upper bound"
g = bsxfun(#ge,X,q(i,:));
l = bsxfun(#lt,X,q(i+1,:));
else % "For the final level we include the upper bound"
g = bsxfun(#ge,X,q(i,:));
l = bsxfun(#le,X,q(i+1,:));
end
Y(g & l) = i;
end
Is there anything I can do to speed this up? Can the code be vectorized?
If I understand correctly, you want to know how many items fell in each bucket.
Use:
n = hist(Y,nbins)
Though I am not sure that it will help in the speedup. It is just cleaner this way.
Edit : Following the comment:
You can use the second output parameter of histc
[n,bin] = histc(...) also returns an index matrix bin. If x is a vector, n(k) = >sum(bin==k). bin is zero for out of range values. If x is an M-by-N matrix, then
How About this
function [Y q] = levels(X,nLevels)
p = linspace(0, 1.0, nLevels+1);
q = quantile(X,p);
Y = zeros(size(X));
for i = 1:numel(q)-1
Y = Y+ X>=q(i);
end
This results in the following:
>>X = [3 1 4 6 7 2];
>>[Y, q] = levels(X,2)
Y =
1 1 2 2 2 1
q =
1 3.5 7
You could also modify the logic line to ensure values are less than the start of the next bin. However, I don't think it is necessary.
I think you shoud use histc
[~,Y] = histc(X,q)
As you can see in matlab's doc:
Description
n = histc(x,edges) counts the number of values in vector x that fall
between the elements in the edges vector (which must contain
monotonically nondecreasing values). n is a length(edges) vector
containing these counts. No elements of x can be complex.
I made a couple of refinements (including one inspired by Aero Engy in another answer) that have resulted in some improvements. To test them out, I created a random matrix of a million rows and 100 columns to run the improved functions on:
>> x = randn(1000000,100);
First, I ran my unmodified code, with the following results:
Note that of the 40 seconds, around 14 of them are spent computing the quantiles - I can't expect to improve this part of the routine (I assume that Mathworks have already optimized it, though I guess that to assume makes an...)
Next, I modified the routine to the following, which should be faster and has the advantage of being fewer lines as well!
function [Y q] = levels(X,nLevels)
p = linspace(0, 1.0, nLevels+1);
q = quantile(X,p);
if isvector(q), q = transpose(q); end
Y = ones(size(X));
for i = 2:nLevels
Y = Y + bsxfun(#ge,X,q(i,:));
end
The profiling results with this code are:
So it is 15 seconds faster, which represents a 150% speedup of the portion of code that is mine, rather than MathWorks.
Finally, following a suggestion of Andrey (again in another answer) I modified the code to use the second output of the histc function, which assigns entries to bins. It doesn't treat the columns independently, so I had to loop over the columns manually, but it seems to be performing really well. Here's the code:
function [Y q] = levels(X,nLevels)
p = linspace(0,1,nLevels+1);
q = quantile(X,p);
if isvector(q), q = transpose(q); end
q(end,:) = 2 * q(end,:);
Y = zeros(size(X));
for k = 1:size(X,2)
[junk Y(:,k)] = histc(X(:,k),q(:,k));
end
And the profiling results:
We now spend only 4.3 seconds in codes outside the quantile function, which is around a 500% speedup over what I wrote originally. I've spent a bit of time writing this answer because I think it's turned into a nice example of how you can use the MATLAB profiler and StackExchange in combination to get much better performance from your code.
I'm happy with this result, although of course I'll continue to be pleased to hear other answers. At this stage the main performance increase will come from increasing the performance of the part of the code that currently calls quantile. I can't see how to do this immediately, but maybe someone else here can. Thanks again!
You can sort the columns and divide+round the inverse indexes:
function Y = levels(X,nLevels)
% "Assign each of the elements of X to an integer-valued level"
[S,IX]=sort(X);
[grid1,grid2]=ndgrid(1:size(IX,1),1:size(IX,2));
invIX=zeros(size(X));
invIX(sub2ind(size(X),IX(:),grid2(:)))=grid1;
Y=ceil(invIX/size(X,1)*nLevels);
Or you can use tiedrank:
function Y = levels(X,nLevels)
% "Assign each of the elements of X to an integer-valued level"
R=tiedrank(X);
Y=ceil(R/size(X,1)*nLevels);
Surprisingly, both these solutions are slightly slower than the quantile+histc solution.